CN109840308A - A kind of region wind power probability forecast method and system - Google Patents

Abstract

The invention provides a method and system for probabilistic forecasting of regional wind power, comprising: collecting the forecast power of a wind farm at a target time, and selecting a forecast power level of the wind farm at the target time from a set of simulated samples obtained based on a pre-built joint probability distribution model The conditional sample set is obtained; a conditional probability distribution function is obtained by fitting the conditional sample set; and a probability prediction interval and a quantile prediction set are extracted based on the conditional probability distribution function. The technical scheme provided by the present invention, according to the established joint probability distribution model, extracts the conditional sample set that meets the wind power probability prediction conditions, and constructs the conditional probability distribution function according to the conditional sample set, which greatly reduces the calculation difficulty and improves the work efficiency.

Description

A method and system for probabilistic forecasting of regional wind power

technical field

The invention belongs to the field of new energy power generation, and in particular relates to a method and system for probabilistic forecasting of regional wind power.

Background technique

The fluctuating and intermittent characteristics of wind energy resources lead to limited prediction accuracy. Therefore, it is necessary to consider the uncertainty distribution of wind power variables in the decision-making of power market and power dispatch to obtain more economical and reasonable results. Therefore, probabilistic forecasting methods that can reflect the uncertainty of power forecasting have been extensively studied. When the research object is the forecast power of multiple wind farms in the region, there is a complex correlation between the random variables in the high-dimensional random vector. How to accurately fit this correlation structure is related to the fitting of the multivariate distribution of the random vector. The combined effect will affect the accuracy of the extracted conditional probability forecast.

The traditional correlation modeling method is constructed by Gaussian Copula model, but the fitting function selected is single, and the modeling accuracy of complex correlation is insufficient.

The traditional probabilistic forecast method, when carrying out the regional total power probabilistic forecast, has a limited number of samples that meet the conditions of the target point, and the effect of the probabilistic forecast is not good.

SUMMARY OF THE INVENTION

The invention selects the R-vine Copula function for correlation modeling, improves the modeling accuracy, and obtains a joint probability distribution model according to the correlation model, which is used in the probability prediction of wind farm power to improve work efficiency.

A method for probabilistic forecasting of regional wind power provided by the present invention includes:

Collect the forecast power of the wind farm at the target time;

From the simulated sample set based on the pre-built joint probability distribution model, select the conditional sample set that meets the forecast power level of the wind farm at the target time;

Fitting the conditional sample set to obtain a conditional probability distribution function;

Probability forecast intervals and quantile forecast sets are extracted based on the conditional probability distribution function.

The construction of the joint probability distribution model includes:

Build a random vector based on the historical data of the wind farm;

Perform marginal distribution fitting on the random variables of the random vector to obtain the marginal cumulative distribution function;

Obtain the correlation vector according to the marginal cumulative distribution function and the random vector;

According to the correlation vector, determine the R-vine copula model;

The joint probability distribution model obtained according to the R-vine copula model and the marginal cumulative distribution function of each random variable;

The historical data includes: historical forecast power and historical forecast error.

The random vector is constructed based on the historical data of the wind farm, including:

A matrix including data at t moments is constructed by taking the data at the same time in the historical data as a row, and the matrix is represented by a random vector.

The R-vine copula model is determined according to the correlation vector, including:

Calculate the Kendall rank correlation coefficient between the two variables in the correlation vector;

Choose a spanning tree structure that maximizes the sum of Kendall rank correlation coefficients;

Determine a binary copula function and perform parameter estimation for each edge in the spanning tree.

The simulated sample set obtained based on the pre-built joint probability distribution model includes:

Arbitrarily generate independent random vectors that satisfy uniform distribution;

Generate a random vector of correlation according to the independent random vector in combination with the R-vine Copula model;

According to the inverse function of the marginal cumulative distribution function, the target random vector is obtained from the random vector of the correlation, and the target random vector is used as a simulated sample set.

Both the fitting of the conditional sample set and the marginal distribution fitting of the random variable of the random vector use the method of kernel density estimation.

A regional wind power probabilistic forecast system provided by the present invention includes:

A model building module for pre-constructing a joint probability distribution model;

The acquisition module is used to collect the forecast power of the wind farm at the target time;

The conditional sample module is used to select the conditional sample set that meets the forecast power of the wind farm at the target time from the simulated sample set obtained based on the pre-built joint probability distribution model;

a fitting module for fitting the conditional sample set to obtain a conditional probability distribution function;

A forecasting module, configured to extract a probability forecast interval and a quantile forecast set based on the conditional probability distribution function.

The model building module includes:

The random vector unit is used to construct a random vector based on the historical data of the wind farm;

The marginal distribution fitting unit is used to perform marginal distribution fitting on the random variables of the random vector to obtain the marginal cumulative distribution function;

The correlation vector unit is used to obtain the correlation vector according to the edge cumulative distribution function and the random vector;

The R-vine copula model unit is used to determine the R-vine copula model according to the correlation vector;

The joint probability distribution model unit is used for the joint probability distribution model obtained according to the R-vine copula model and the marginal cumulative distribution function of each random variable;

The historical data includes: historical forecast power and historical forecast error.

The conditional sample module includes:

The first generating unit is used to arbitrarily generate independent random vectors satisfying uniform distribution;

a second generating unit, configured to generate a random vector of correlation according to the independent random vector in combination with the R-vine Copula model;

a sample determination unit, configured to obtain a target random vector from the random vector of the correlation according to the inverse function of the edge cumulative distribution function, and use the target random vector as a simulated sample set;

The screening unit is used for screening out the conditional sample set that meets the forecast power of the wind farm at the target time from the simulation sample set.

The R-vine copula model unit includes:

The coefficient calculation subunit is used to calculate the Kendall rank correlation coefficient between the two variables in the correlation vector;

The spanning tree subunit is used to select a spanning tree structure that maximizes the sum of Kendall rank correlation coefficients;

The model determination subunit is used to determine the binary copula function and perform parameter estimation for each edge in the spanning tree.

Compared with the closest prior art, the technical solution provided by the present invention has the following beneficial effects:

According to the technical scheme provided by the present invention, according to the establishment of a joint probability distribution model, a conditional sample set that satisfies the probability prediction conditions of wind power power is extracted, and a conditional probability distribution function is constructed according to the conditional sample set, which greatly reduces the calculation difficulty and improves the work efficiency;

In the technical scheme provided by the present invention, the R-vine Copula function can be used to realize the splitting of high-dimensional correlation structures and the selection and fitting of various binary Copula functions, thereby improving the flexibility and accuracy of correlation modeling.

Description of drawings

1 is a flow chart of a method for probabilistic forecasting of regional wind power according to the present invention;

2 is an overall flow chart of a method for probabilistic forecasting of regional wind power according to an embodiment of the present invention.

Detailed ways

Below in conjunction with accompanying drawing, the present invention is described in further detail:

Example 1:

Fig. 1 is a flow chart of a method for probabilistic forecasting of regional wind power according to the present invention. As shown in Fig. 1, a method for probabilistic forecasting of regional wind power provided by the present invention may include:

Collect the forecast power of the wind farm at the target time;

From the simulated sample set based on the pre-built joint probability distribution model, select the conditional sample set that meets the forecast power level of the wind farm at the target time;

Fitting the conditional sample set to obtain a conditional probability distribution function;

Probability forecast intervals and quantile forecast sets are extracted based on the conditional probability distribution function.

The construction of the joint probability distribution model includes:

Build a random vector based on the historical data of the wind farm;

Perform marginal distribution fitting on the random variables of the random vector to obtain the marginal cumulative distribution function;

Obtain the correlation vector according to the marginal cumulative distribution function and the random vector;

According to the correlation vector, determine the R-vine copula model;

The joint probability distribution model obtained according to the R-vine copula model and the marginal cumulative distribution function of each random variable;

The historical data includes: historical forecast power and historical forecast error.

The random vector is constructed based on the historical data of the wind farm, including:

A matrix including data at t moments is constructed by taking the data at the same time in the historical data as a row, and the matrix is represented by a random vector.

The R-vine copula model is determined according to the correlation vector, including:

Calculate the Kendall rank correlation coefficient between the two variables in the correlation vector;

Choose a spanning tree structure that maximizes the sum of Kendall rank correlation coefficients;

Determine a binary copula function and perform parameter estimation for each edge in the spanning tree.

The simulated sample set obtained based on the pre-built joint probability distribution model includes:

Arbitrarily generate an independent random vector that satisfies a uniform distribution;

Generate a random vector of correlation according to the independent random vector in combination with the R-vine Copula model;

According to the inverse function of the marginal cumulative distribution function, the target random vector is obtained from the random vector of the correlation, and the target random vector is used as a simulated sample set.

Both the fitting of the conditional sample set and the marginal distribution fitting of the random variable of the random vector use the method of kernel density estimation.

Embodiment 2:

Based on the same inventive concept, a regional wind power probabilistic forecast system provided by the present invention may include:

A model building module for pre-constructing a joint probability distribution model;

The acquisition module is used to collect the forecast power of the wind farm at the target time;

The conditional sample module is used to select the conditional sample set that meets the forecast power of the wind farm at the target time from the simulated sample set obtained based on the pre-built joint probability distribution model;

a fitting module for fitting the conditional sample set to obtain a conditional probability distribution function;

A forecasting module, configured to extract a probability forecast interval and a quantile forecast set based on the conditional probability distribution function.

The model building module includes:

The random vector unit is used to construct a random vector based on the historical data of the wind farm;

The marginal distribution fitting unit is used to perform marginal distribution fitting on the random variables of the random vector to obtain the marginal cumulative distribution function;

The correlation vector unit is used to obtain the correlation vector according to the edge cumulative distribution function and the random vector;

The R-vine copula model unit is used to determine the R-vine copula model according to the correlation vector;

The joint probability distribution model unit is used for the joint probability distribution model obtained according to the R-vine copula model and the marginal cumulative distribution function of each random variable;

The historical data includes: historical forecast power and historical forecast error.

The conditional sample module includes:

The first generating unit is used to arbitrarily generate independent random vectors satisfying uniform distribution;

a second generating unit, configured to generate a random vector of correlation according to the independent random vector in combination with the R-vine Copula model;

a sample determination unit, configured to obtain a target random vector from the random vector of the correlation according to the inverse function of the edge cumulative distribution function, and use the target random vector as a simulated sample set;

The screening unit is used for screening out the conditional sample set that meets the forecast power of the wind farm at the target time from the simulation sample set.

The R-vine copula model unit includes:

The coefficient calculation subunit is used to calculate the Kendall rank correlation coefficient between the two variables in the correlation vector;

The spanning tree subunit is used to select a spanning tree structure that maximizes the sum of Kendall rank correlation coefficients;

The model determination subunit is used to determine the binary copula function and perform parameter estimation for each edge in the spanning tree.

The random vector unit includes:

A matrix including data at t moments is constructed with the data at the same time as a row, and the matrix is represented by a random vector.

Embodiment three:

A method for probabilistic forecasting of regional wind power, which can include:

Build an R-vine copula model and make probabilistic forecast of wind power based on the R-vine copula model;

Figure 2 is an overall flow chart of a method for probabilistic forecasting of regional wind power. As shown in Figure 2, the construction of the R-vinecopula model may include:

Step 1-1: Input d-dimensional samples [X ₁ ,...,X _d ]=[P ₁ ,...,P _n ,E ₁ ,...,E _n ];

Step 1-2: Perform edge fitting on the elements in the sample to obtain the fitted edge cumulative distribution function;

Step 1-3: remove the influence of marginal distribution in [X ₁ ,...,X _d ] to obtain d-dimensional sample data [U ₁ ,...,U _d ];

Step 1-4: Based on the d-dimensional sample data [U ₁ ,...,U _d ], filter out the maximum spanning tree according to the principle of maximizing the correlation coefficient, and obtain the relevant variable pairs in the tree; check whether the relevant variable pairs are independent, if yes Determine whether all trees have been generated, otherwise filter out the binary copula function required by the tree and perform parameter estimation to determine whether all trees have been generated. If all trees are generated, the R-vine copula model construction is completed. For tree generation, repeat this step until all trees are generated.

After all trees are generated, the step of probabilistic forecasting of wind power based on the R-vine copula model may be performed, wherein the probabilistic forecast of wind power based on the R-vine copula model may include:

Step 2-1: Based on the R-vine copula model, obtain the simulated sample set S by random sampling;

Step 2-2: According to the probability forecast conditions, select the samples that meet the forecast target points in S;

Step 2-3: Perform conditional probability density distribution fitting on the selected samples, and output the conditional probability density function.

specific:

Step 1-1: The generation of the d-dimensional samples in the input d-dimensional samples [X ₁ ,...,X _d ]=[P ₁ ,...,P _n ,E ₁ ,...,E _n ] can be include:

The objects that need to be modeled are the forecast power p and forecast error e of each wind farm in the region. (5-1) gives an example of n wind farms, where each row in the matrix corresponds to the data at the same time, according to the size of the sample set There are a total of t time data, and the random variables corresponding to each column are represented by uppercase P and E, which are uniformly represented as a d-dimensional random vector X=(X ₁ ,...,X _d ) for the convenience of expression.

Step 1-2: Perform edge fitting on the elements in the sample to obtain a fitted edge distribution model, which may include:

Considering that it is difficult to determine a uniform parametric distribution for each marginal variable and the empirical distribution function is discontinuous, this method uses the non-parametric distribution of kernel density estimation to fit the marginal distribution, and the fitted probability density function is shown in Equation (5-2) shown.

Among them, h is the bandwidth, K represents the kernel function, and the present invention uses a Gaussian kernel, and the expression is shown in formula (5-3), n represents the sample size, and X represents the sample data.

Step 1-3: Remove the influence of marginal distribution in [X ₁ ,...,X _d ] to obtain d-dimensional sample data [U ₁ ,...,U _d ], which can include:

The corresponding marginal cumulative distribution function according to the estimate and X can obtain the correlation vector U=(U ₁ , . . . , U _d ) stripped of the influence of the edge distribution, and the corresponding edge distribution satisfies the uniform distribution.

Convert X to U through the edge cumulative distribution function, strip the influence of the edge distribution, and only consider the relevant structure such as Eq.

in, is the inverse function of the marginal cumulative distribution function CDF, U=(U ₁ ,...,U _d )∈[0,1] ^d .

Step 1-4: Based on the d-dimensional sample data [U ₁ ,...,U _d ], screen out the maximum spanning tree according to the principle of maximizing the correlation coefficient, and obtain the relevant variable pairs in the tree, and test whether the relevant variable pairs are independent, if yes, then Determine whether all trees have been generated, otherwise filter out the binary copula function required by the tree and perform parameter estimation to determine whether all trees have been generated. If all trees are generated, the R-vine copula model construction is completed. For tree generation, this step is repeated until all trees are generated, which can include:

According to the U obtained in the previous step, determining the corresponding R-vine copula requires the following three tasks:

1. Select the R-vine structure, that is, give the set of constraints {j(e), k(e)D(e)} for each tree.

2. Select a suitable binary copula type for the binary random variable corresponding to each edge e in R-vine.

3. Estimate the parameters of each binary copula function.

The above three items are often combined in practical applications, and the construction of R-vine copula is implemented tree by tree according to the successive method. The algorithm flow is as follows:

Based on the consideration of the computational cost of the model, an independence test is introduced before binary copula fitting and parameter estimation. For pairs of random variables that are close to independent, an independent copula function is directly used to control the computational complexity of the corresponding model according to the degree of significance. degree and precision.

Specifically, the probabilistic forecast of wind power based on the R-vine copula model may include:

According to the R-vine copula model, the joint probability distribution function is obtained. Although the joint probability distribution function is a mathematical expression of continuous analysis, when the regional total power probability forecast is performed, the calculated object needs to be calculated by multiple integrals, and in the integral function In complex cases, it is very difficult to perform integral calculation and has no practical significance in engineering. Therefore, it is considered to generate a sufficient number of data samples according to the obtained joint probability distribution function to fit the conditional probability density results that meet the conditions.

First, a d-dimensional independent random vector W:=(W ₁ ,...,W _d ) satisfying the uniform distribution U(0,1) ^d is randomly generated by the computer.

Then generate a random vector of correlations by (5-4)

Finally, according to the inverse function of the fitted edge cumulative distribution function, namely from Find the target random vector After obtaining a sufficient number of simulated samples (simulation sample set S), it is considered that Contains all the information of the joint distribution, the present invention selects the sample points of the total power of the region that meet the conditions according to the forecast power level of each wind farm in the region, and forms a set C, and then adopts the method of kernel density estimation for the samples in C ( The same formula 5-1) builds a continuous probability distribution function, and extracts forecast interval results with different confidence levels according to actual needs.

The joint probability distribution function is obtained according to the R-vine copula model, which can include:

Regular vine (Regular vine, R-vine) V is a regular vine of d elements, and the set of its edges is expressed as E(V)=E ₁ ∪…∪E _i ∪…∪E _d-1 , where E _i , i=1,...,d-1 represents the set of edges of the i-th tree T _i . Rule vines need to meet the following three conditions:

4) V={T ₁ ,...,T _d-1 }, that is, a set composed of d-1 trees.

5) The node set of T ₁ is N ₁ ={1,...,d}, and the edge set is E ₁ ; and for i=2,...,d-1, the node set of T _i is N _i , and the edge set is E _i , the condition Ni =E _{i -1} needs to be satisfied.

6) (Proximity principle) For i=2,...,d-1, {a,b}∈E _i , #(a△b)=2, where △ represents the equivalent difference of the calculation set, # represents the calculation set potential.

7) The regular vine V with dimension d is composed of d-1 trees {T ₁ ,...,T _d-1 }, and its node set is {N ₁ ,...,N _d-1 }, where the set N ₁ ={ 1,...,d}, the number corresponding to the initial d random variables in the correlation modeling. The edge set of V is represented as {E ₁ ,...,E _d-1 }, and an edge e in the edge set E _i of the tree T _i can be represented as e=j(e),k(e)|D(e) The form of , where {j(e), k(e), j(e)≠k(e)} is called the conditioned set, and D(e) is called the conditioning set, and the elements in these two sets are represented by {1 ,…,d} constitutes. According to the principle of proximity, e is determined by the corresponding two edges a=j(a), k(a)|D(a), b=j(b), k(b)|D(b) in E _i-1 , a and b have a common node in T _i-1 , then the relationship of the three edges is expressed as the following two relationships

D(e):=U(a)∩U(b) (3-4)

{j(e),k(e)}:=U(a)∪U(b)\D(e), (3-5)

Among them, U(e):={j(e), k(e), D(e)} represents the complete set of elements contained in e, including all elements in the conditioning set and the conditioned set. In addition, for the edge in E1, its form is e ₌ j(e), k(e), because the conditioning set D(e) at this time is an empty set.

When the edge labeling rules are clarified, the binary copula density corresponding to e can be expressed as c _{j(e),k(e)|D(e)} .

Combined with the above content, the formula of the multi-dimensional joint density distribution described by the regular vine structure is given:

Its corresponding d-dimensional random variable X:=(X ₁ ,...,X _d ), the marginal cumulative distribution function is f _k , k=1,...,d, X _D(e) represents X by D(e ) specified subset. Among them, the variables in a binary copula in the ith tree in the formula - conditional distribution functions F(x _j(e) |x _D(e) ) and F(x _k(e) |x _{D(e )} ) can be calculated by the copula function C and the corresponding conditional distribution function F in the i-1th tree whose parameters have been estimated.

As will be appreciated by those skilled in the art, the embodiments of the present application may be provided as a method, a system, or a computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It will be understood that each flow and/or block in the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to the processor of a general purpose computer, special purpose computer, embedded processor or other programmable data processing device to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing device produce Means for implementing the functions specified in a flow or flow of a flowchart and/or a block or blocks of a block diagram.

These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory result in an article of manufacture comprising instruction means, the instructions The apparatus implements the functions specified in the flow or flow of the flowcharts and/or the block or blocks of the block diagrams.

These computer program instructions can also be loaded on a computer or other programmable data processing device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process such that The instructions provide steps for implementing the functions specified in the flow or blocks of the flowcharts and/or the block or blocks of the block diagrams.

Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit its protection scope. Although the application has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: After reading this application, those skilled in the art can still make various changes, modifications or equivalent replacements to the specific embodiments of the application, but these changes, modifications or equivalent replacements are all within the protection scope of the pending claims.

Claims (10)

1. A regional wind power probability forecasting method is characterized by comprising the following steps:

acquiring forecast power of a wind power plant at a target moment;

screening out a condition sample set which accords with the forecast power level of the wind power plant at a target moment from a simulation sample set obtained based on a pre-constructed joint probability distribution model;

fitting the conditional sample set to obtain a conditional probability distribution function;

and extracting a probability forecast interval and a quantile forecast set based on the conditional probability distribution function.

2. The regional wind power probability forecasting method of claim 1, wherein the building of the joint probability distribution model comprises:

constructing a random vector based on historical data of the wind power plant;

performing edge distribution fitting on random variables of the random vectors to obtain an edge cumulative distribution function;

obtaining a correlation vector according to the edge cumulative distribution function and the random vector;

determining an R-vine copula model according to the relevance vector;

obtaining a joint probability distribution model according to the R-vine copula model and the edge cumulative distribution function of each random variable;

the historical data includes: historical forecast power and historical forecast error.

3. The regional wind power probability forecasting method of claim 2, characterized in that the construction of the random vector based on the historical data of the wind farm comprises:

and constructing a matrix comprising t time data by taking the data at the same time in the historical data as a row, and representing the matrix by using a random vector.

4. The regional wind power probability forecasting method of claim 2, wherein the determining an R-vine copula model according to the relevance vector comprises:

calculating Kendall rank correlation coefficients between every two variables in the correlation vector;

selecting a spanning tree structure which meets the Kendall rank correlation coefficient sum maximization;

a binary copula function is determined for each edge in the spanning tree and parameter estimation is performed.

5. The regional wind power probability forecasting method of claim 2, wherein the obtaining of the simulation sample set based on the pre-constructed joint probability distribution model comprises:

generating independent random vectors meeting the uniform distribution at will;

generating a random vector of correlation according to the independent random vector and the R-vine Copula model;

and obtaining a target random vector from the random vector of the correlation according to the inverse function of the edge cumulative distribution function, and taking the target random vector as a simulation sample set.

6. The regional wind power probability forecasting method of claim 2, characterized in that fitting the condition sample set and fitting the edge distribution of the random variables of the random vector both adopt a kernel density estimation method.

7. A regional wind power probability forecasting system is characterized by comprising:

the model construction module is used for constructing a joint probability distribution model in advance;

the collection module is used for collecting the forecast power of the wind power plant at a target moment;

the system comprises a conditional sample module, a power source module and a power source module, wherein the conditional sample module is used for screening out a conditional sample set which accords with the forecast power of a wind power plant at a target moment from a simulated sample set obtained based on a pre-constructed joint probability distribution model;

the fitting module is used for fitting the conditional sample set to obtain a conditional probability distribution function;

and the forecasting module is used for extracting a probability forecasting interval and a quantile forecasting set based on the conditional probability distribution function.

8. The regional wind power probability forecasting system of claim 7, wherein the model building module comprises:

the random vector unit is used for constructing a random vector based on historical data of the wind power plant;

the edge distribution fitting unit is used for performing edge distribution fitting on the random variable of the random vector to obtain an edge cumulative distribution function;

the correlation vector unit is used for obtaining a correlation vector according to the edge cumulative distribution function and the random vector;

the R-vine copula model unit is used for determining an R-vine copula model according to the relevance vector;

the joint probability distribution model unit is used for obtaining a joint probability distribution model according to the R-vine copula model and the edge cumulative distribution function of each random variable;

the historical data includes: historical forecast power and historical forecast error.

9. The regional wind power probability forecasting system of claim 8, wherein the condition sample module comprises:

a first generating unit for arbitrarily generating independent random vectors satisfying uniform distribution;

the second generation unit is used for generating a random vector of correlation according to the independent random vector and an R-vine Copula model;

the sample determining unit is used for obtaining a target random vector from the random vector of the correlation according to an inverse function of the edge cumulative distribution function, and the target random vector is taken as a simulation sample set;

and the screening unit is used for screening out a condition sample set which accords with the forecast power of the wind power plant at the target moment from the simulation sample set.

10. The regional wind power probability forecasting system of claim 8, wherein the R-vine copula model unit comprises:

the coefficient calculating subunit is used for calculating Kendall rank correlation coefficients between every two variables in the correlation vector;

the spanning tree subunit is used for selecting a spanning tree structure which meets the Kendall rank correlation coefficient sum maximization;

and the model determining subunit is used for determining a binary copula function for each edge in the spanning tree and performing parameter estimation.